DICOM @ OFFIS

Is there a need for secure communication using SChannel instead of OpenSSL in DCMTK on Windows? On the Windows platform, the OpenSSL library has two main issues: 1. OpenSSL is incompatible with secure key storage on TPM chips on the Windows platform. Having this capability would improve security in ...

Currently, I am unable to build dcmdata from the master branch on VS 2019 due to a seemingly missing include of Windows.h

See https://github.com/DCMTK/dcmtk/pull/72 for a fix.

In my setup I build with OpenSSL and ICU. Not sure if this has anything to do with it.

Great news! Thank you Micahael.

Jøger

Hi Michael, The general issue is that it is impossible for a derived class to reimplement DcmSCU::sendMOVERequest or any of the other sendXXXRequest functions as long as the nextMessageID function is private. The specific issue, is that we want to get hold of the message ID before the C-MOVE request...

Hi again Michael, I once again stumbled across a problem caused by the private DcmSCU::nextMessageID function. DcmSCU appears to be designed with extensibility through inheritance in mind. If this is the intent, I believe that the nextMessageID must be made protected instead of private, as there are...

Thank you Michael and Vasyl. I work in Vasyl's team. I fully agree with the principle Michael, and I too understand that it is a value for DCMTK and OFFIS to see the DCMTK identification. We will also be working within our organization to understand if it can be acceptable to use the OFFIS implement...

This kind of problem can show up if the dcmtkcrypto_d.lib, dcmtkcrypto_o.lib, dcmtkssl_d.lib, or dcmtkssl_o.lib are not present in the expected directory when configuring the project. I assume the heuristics for determining if OpenSsl can be found does not check presence of all libs. So, after deter...

I am building DCMTK (static libs) with ICU and OpenSSL on Windows with Visual Studio. To reduce build times of client software on CI/CD and developer computers, I would like to create a Nuget package (archive) that is downloaded from an artifact server. This reduces build times significantly compare...

Thank you so much! I hope for the best, and look forward to enjoying the N-SET experience.

Hi Michael, When you have time, could you please have a look at https://github.com/DCMTK/dcmtk/pull/63? It is an attempt to prevent port conflicts when using DcmSCP in unit tests. Is it correct that tests marked as EF_Slow are not run on the build servers? Maybe this could explain why the older dcmn...

Thank you Micahel, this sounds like a good explanation for the problems. I assume this is a universal challenge for automated testing of DICOM networking functionality. Maybe it could be possible to let the OS assign a free port if the DcmSCP is configured with port 0. DcmSCP seems to already have a...

Hi again Michael, I am trying to understand why the unit tests failed in your build environment. I believe that if these kind of tests fail on a build server, it is not unlikely that the code could fail also in a real life scenario. Such tests are extremely useful for detecting hard-to-reproduce bug...

Thank you for the input! https://github.com/DCMTK/dcmtk/pull/62 shows a preliminary sketch which could be one way to be able to cancel ongoing association attempts while the DcmSCU is attempting to establish a TCP connection to the unreachable server. It allows waking the dulfsm.cc requestAssociatio...

Hi OFFIS team, I just created https://github.com/DCMTK/dcmtk/pull/61. This change makes it possible to configure the DcmSCU TCP connection timeout per instance, instead of using the global dcmConnectionTimeout parameter. In a multi-service executable, that for example supports QR, Worklist, and Stor...

Hi Michael, https://github.com/DCMTK/dcmtk/pull/57 (Extend DcmScu with support for N-SET requests) is rebased on top of master, and with some slight fixes since the previous draft. In particular, the attributeList of the DcmSCU::sendNSETRequest is made mandatory. This also goes for DcmSCU::sendNCREA...